Upload
lamdan
View
215
Download
0
Embed Size (px)
Citation preview
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-1
CHAPTER TWO
INFORMATION TECHNOLOGY
2.1 INTRODUCTION
Rapid developments in information technology (IT), which is defined as both
computing and telecommunications, have affected all aspects of business, not just supply
chain management. During the 1990’s, sales of IT grew to the extent that it has become
the largest industry in the United States exceeding construction, food products and
automotive manufacturing.1 As shown in Figure 2.1, business investment in hardware and
software has increased exponentially since the mid-1980’s, reaching a torrid pace by the
end of the 1990’s, with no end in sight.2 Moreover, these investments do not include sky-
rocketing fees paid for IT consulting services.
The creation and management of corporate databases has been facilitated by
widespread implementation of Enterprise Resource Planning (ERP) systems. These
systems offer the promise of transactional data bases that are standardized across the
company, thereby facilitating integration of supply chain activities. Models are playing
an increasing role in helping managers extract effective supply chain plans from these
databases. Nevertheless, the purposes of such models, and their potential for improving a
company’s competitive advantage, are not yet fully understood by many supply chain
managers.
In §2.2, we provide a brief overview of developments in ERP and e-commerce
systems from the perspective of supply chain management. These systems are primarily
Transactional IT concerned with acquiring, processing and communicating raw data about
the company’s supply chain, and with the compilation and dissemination of
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-2
Bu
sin
ess
Inv
estm
ent
in C
om
pu
ter
Har
dw
are
and
So
ftw
are
Fig
ure
2.1
Sou
rce:
Lo
hr
[199
9]
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-3
reports summarizing these data. They also facilitate communication among different
companies in an extended or virtual supply chain. In §2.3, we compare Transactional IT
with Analytical IT, which is created to assist managers in making supply chain decisions.
Analytical IT employs models constructed from supply chain decision databases that
are derived from the company’s transactional databases. Analytical IT is comprised of
these supply chain decision databases, plus modeling systems and programs linking
corporate databases to the supply chain decision databases.
In theory, the modeling systems in a company’s Analytical IT and their decision
databases should be organized in a suite of inter-connected applications for strategic,
tactical and operational planning. This hierarchy of modeling systems is discussed in
detail in §2.4. In practice, very few companies have yet come close to achieving a
complete suite. Nevertheless, we believe there is value in laying out a comprehensive
structure for these modeling systems and discussing linkages among the various
components.
In §2.5, we discuss issues connected with enhancing legacy systems for and legacy
thinking about supply chain modeling. These issues are a preview of behavioral research,
which we discuss in Chapter 12, into barriers inhibiting rational decision making using
models. The chapter concludes with §2.6 where we present final thoughts about the
information revolution, especially as it relates to supply chain management.
2.2 DEVELOPMENTS INENTERPRISE RESOURCE PLANNING SYSTEMS
AND ELECTRONIC COMMERCE
The development of information technology (IT) for managing and communicating
transactional data has been a primary focus of computer scientists and information
technologists for over 40 years.3 Figure 2.1 reveals the great expansion during the 1990’s
of managerial interest and investment in hardware and software for these purposes. Still,
even with the advent of Enterprise Resource Planning (ERP) systems and e-commerce,
we would be naïve to suppose that companies have achieved permanent solutions to their
data management problems.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-4
It is beyond the aim of this book to discuss these software and hardware
developments in great detail. Rather, we will be concerned in this chapter with the synergy
between IT and supply chain management using modeling systems.4 Today’s IT enables
the development of such systems, but paths to their successful implementation remain
unclear in most companies. Although scholars and consultants often use expressions such
as “the company must (can, should) optimize supply chain decisions,” details about how
such analysis can be performed are often missing. Moreover, managers do not yet fully
realize how models can provide a comprehensive, high level view of the tangled forest of
their transactional supply chain data. Based on their growing interest in modeling systems,
we expect managers will soon become much better informed. By participating in processes
to extract, aggregate, extrapolate, and otherwise transform transactional data into input
data required by modeling systems to support decision making, managers will gain a fresh
and important perspective on how to exploit IT advances.
In this section, we review recent developments in ERP systems and e-commerce.
Our discussion of e-commerce is divided into separate examinations of business-to-
consumer and business-to-business e-commerce. Business-to-business linkages via the
Internet require extended or new ERP systems that facilitate communication among
companies of diverse sizes and missions.
ERP SYSTEMS
An ERP system is comprised of software and hardware that facilitates the flow of
transactional data in a company relating to manufacturing, logistics, finance, sales and
human resources. In principle, all business applications of the company are integrated in
an uniform system environment that accesses a centralized database residing on a common
platform. Common and compatible data fields and formats are used across the entire
enterprise. Moreover, data is entered once and only once ensuring that all applications
make consistent use of these data.
In 1997, worldwide sales of ERP systems and related services were estimated to
exceed $10 billion. Of this total, approximately 50% was for software, 30% for
installation, training, customization, and 20% for maintenance and upgrades.5 The large
percentage and absolute amount spent on installation, training and customization reflects
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-5
the need for significant business process re-design to bring systems and practices across
diverse functions into alignment. In addition to improving human data collection and
communication processes, re-engineering should, but does not always, include cleansing
and tightening of data definitions: for example, a reduction in the number of SKU’s to
those that are truly distinct.
Implementing ERP systems has proven unexpectedly difficult in many
companies. Numerous articles have appeared in magazines and journals offering guidelines
for avoiding the worst headaches. Even when an implementation project is well managed,
the result can be disappointing due to inherent limitations of current ERP systems. These
include6
• Imposed conformity – the ERP system imposes rigid requirements on data and
processes that often inhibit the way the company can operate its business
• Inability to employ software from multiple vendors – the company cannot integrate
modules, including modeling systems, from multiple vendors with the monolithic ERP
system acquired from the primary vendor
• Incompatibility of ERP systems across the supply chain – the company cannot easily
integrate supply chain databases with vendors and customers, especially those who
are too small to afford a massive ERP implementation
Current thinking is that these problems will be overcome by new ERP systems that are
modular and Web-enabled.7 Individual modules for transactional data management and
modeling analysis, often developed by third-party vendors, will be bolted onto ERP
systems using middleware, which provides standard interfaces for the modules.
The rapid growth in e-commerce has magnified these deficiencies of ERP systems.
The expectation is that Internet-driven re-engineering will require integrating business
processes across corporate boundaries. Modular, flexible ERP systems will be essential
for implementing inter-enterprise information systems for supply chains comprised of
several companies of varying sizes and cultures. Moreover, since e-commerce is so new,
Internet companies will need the capability to modify their ERP systems as new
conditions emerge.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-6
E-COMMERCE
Communication over the Internet is characterized by easy accessibility, low cost
to use, and speed. These features are the result of astonishing developments in recent
years in computer networks, processing speed, data storage capabilities, software,
display technology, and user interfaces. The linking of people and companies due to e-
commerce has opened up exciting new marketing opportunities as well as new processes
for improving supply chain management. Direct business-to-consumer marketing over the
Internet of products such as clothes, groceries, and PCs is an emerging concept that has
become red hot. Business-to-business communication over the Internet is also expanding
rapidly. We discuss these two new developments in the paragraphs that follow. We
conclude by examining Internet systems and processes that facilitate the creation of
supply and demand markets for commodities.
Business-to-Consumer E-Commerce
Business-to-consumer e-commerce is a new method of retailing that puts the
consumer in direct contact with a grocery, clothing, or PC company offering products.
These Internet companies faces a host of new marketing and sales challenges including
• devising graphics to attractively display physical products on the website
• pricing products to gain market share, reflect supply chain costs, or some other
criteria that change radically with evolving markets
• extrapolating sales patterns from initial markets to new markets
• identifying demographics of website customers
• devising acceptable and sustainable customer service criteria
• devising strategies to retain customers
• selecting the number and range of products to offer that the website and the
supply chain can support
• connecting website sales to physical inventory
• providing security for payment by credit cards
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-7
Innovations of business-to-customer Internet companies occur largely on the demand side
of their businesses, although changes to conventional thinking are also needed for supply
chain management.
Despite the excitement of business-to-customer Internet companies, their sales are
still only a small percentage of total retailing sales. Moreover, they are not projected to
make serious inroads in the near future in most industries. For example, the market for
home delivery of groceries is forecast to reach $3.5 billion by 2002, which represents only
0.7% of total grocery sales projected for that year. Similarly, clothes and accessories
acquired over the Internet are forecast to represent only 1.6% of total industry sales in
2002.8
The logistics of business-to-customer Internet companies is driven by the classical
order fulfillment principle: Deliver the correct product to the correct location at the
correct time for a competitive price. These companies have only just begun to realize the
full importance of this principle, and how difficult it is to effectively link their web-based
marketing and sales activities to their order fulfillment activities. In simple terms, an
Internet company has two choices. They can build their own warehouses and manage
their own distribution systems or they can hire third party logistics companies to handle
distribution for them.
Either option can prove costly. An Internet company can expect to spend
between $60 Million to $80 Million in constructing a one million square foot warehouse,
a size that is needed for a high volume company, especially during peak periods. In
addition to needing access to large amounts of capital, the company obviously has to be
confident that they will sustain a justifiable level of sales volume over the foreseeable
future. Alternatively, they can employ a third party distributor who will charge them
roughly 10 percent of gross sales to fulfill their orders.9
An additional problem for a business-to-consumer Internet company is the need
to link their website with the system that manages their inventory. This connection is
critical if customers are to receive real-time information about product availability. Again,
if the Internet company has sufficient capital, in addition to building warehouses, it can
develop customized integrated systems with seamless connections among the website, the
inventory management system, and other systems needed to manage the company’s
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-8
supply chain. Such an integrated system is discussed in §10.4 where we examine supply
chain management and modeling systems in a company making home deliveries of
groceries. If, however, the Internet company decides to employ a third party distributor,
it runs the risk of trying to link its website and order entry system with the distributor’s
software for inventory management and tracking. Incompatibilities with the distributor’s
systems might cause headaches as large as those experienced in developing a customized
integrated system.
Finally, an Internet company must anticipate escalating demands from customers
for higher levels of service. For example, customers seeking home delivery of groceries
will prefer companies offering same day delivery with tight time windows over those
offering next day delivery with looser time windows. For manufactured products such as
PC’s, customers will prefer companies offering to more fully customize their purchases
over those offering little or no customization. To meet these increasing pressures, faster
and more powerful data management and modeling systems are required.
In summary, supply chain management of business-to-consumer Internet
companies is subject to serious economies of scale in paybacks from investments in
warehouses, inventories, and integrated data management and modeling systems. Due to
high and increasing customer service expectations, these economies of scale may be even
more pronounced than those experienced by traditional retailers. It suggests that we may
soon see a serious reduction in the number of Internet companies in a given industry as
mergers and acquisitions allow companies to achieve volumes justifying capital
investment in brick and mortar and integrated systems.
Business-to-Business E-Commerce
Although investors are enthralled with the promise of business-to-consumer e-
commerce, the potential impact of business-to-business e-commerce on supply chain
management is much larger. In principle, business-to-business systems can provide
effective inter-enterprise communication of data and plans for manufacturing and
distribution across virtual supply chains in many industries. This might allow companies
to move work across corporate boundaries, reduce cycle times by direct interconnection,
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-9
and develop collaborative forecasts. But many barriers must be overcome if the promise
of virtual supply chains is to be realized.
First, multiple companies operating a virtual supply chain must have standardized
definitions and meanings of data. A decision to shift production of a product from one
company to a second in the virtual supply chain because the unit production cost is lower
in the second company assumes that this cost is consistently defined and measured across
the two companies. Moreover, systems must be in place for seamless integration of data.
Such requirements are, of course, no different than those addressed and allegedly met by
ERP systems. And, as we discussed above, ERP companies are currently active in
developing modules and middleware that they believe will allow standardization and
efficient communication to be achieved across virtual supply chains.
Second, virtual supply chains assume a level of inter-company coordination that is
often not achieved today among business units of the same company. For example, we
cannot presume that a manufacturer of consumer durables will be willing to share
sensitive cost data with a major OEM distributor of these products, especially when the
manufacturer sells to other distributors and to its own franchised stores. Another issue is
how companies in a virtual supply chain will agree to split cost savings realized from
improved business-to-business communication and supply chain management.
Third, faster communication of data does not automatically lead to better decision
making. As we discuss in the following sections of this chapter, competitive supply chain
management cannot be achieved merely by rapid and myopic response to today’s supply
chain needs. Optimization modeling systems are required to unravel the complex
interactions and ripple effects that make supply chain management difficult and
important.
Procurements over the Internet, Spot Markets and Auctions
A variety of electronic markets have emerged over the past five years where
commodities, collectibles, and other products are bought and sold. Our interest is
primarily in business-to-business sales over the Internet, which could exceed $66 billion
during the year 2000.10 This development has been called Internet-based procurement
(IBP). It is concerned with direct procurement of specific parts and components needed
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-10
for production by buyers, and indirect procurement of commodities and products not
customized for production such as generic parts, office supplies, maintenance, and so on.
Electronic procurement is developing along three dimensions11
• Seller-side sites – Suppliers place their catalogs and spec sheets on their Web sites.
The functionality of the sites may be expanded to include search capabilities and
electronic commerce functionality such as credit card sales. Boise Cascade and Office
Depot have sites of this type.
• Buyer-side sites – Buyers have installed software, which is available from software
providers, allowing them to read and standardize vendor catalogs. The software also
allows the creation, routing, approval and submission of orders.
• Third-party sites – These are neutral sites that serve as marketplaces where buyers
and sellers can link up. They provide catalog information from several sellers and
usually capabilities supporting direct sales between the buyer and the seller. These
sites are usually specific to certain industries such as specialty chemicals or
pharmaceuticals.
Most of the transactions performed on these sites are indirect procurements of
commodity products. Market researchers have found that Internet purchasing has led to
improved corporate-wide purchasing strategies, lower transaction costs, and improved
productivity among buyers.
Direct procurement over the Internet by manufacturing firms can be more
complicated because the required parts and components may require some customization.
Moreover, factors beyond cost, such as quality, on-time delivery, and supplier flexibility,
may be very important. Richer, more flexible software solutions are needed, but they
require customization and can be very expensive. For this reason, some industries are
seeking to establish common standards for their suppliers. For example, a trade
association in the automotive industry commissioned the implementation of a standards-
based network, the Automotive Network eXchange (ANX). Recently, Ford and General
Motors announced that their suppliers will soon be required to be connected to and use
this network. The expectation is that communication over ANX with suppliers in the
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-11
second, third and fourth tiers will take hours, rather than weeks using the current
hodgepodge of systems.
A related development is electronic markets where buyers and sellers participate
in auctions of goods and services. The intention of these auctions is to enhance economic
efficiency through aggregation and matching.12 Aggregation refers to the assembly of very
large numbers of buyers and sellers. Matching refers to dynamic processes whereby
buyers are able to link up with sellers offering products that best match their needs.
These developments suggest the possible emergence of spot markets for industrial
products and services, including contract manufacturing and shipping, that will seriously
affect supply chain strategies in many industries.
2.3 COMPARISON OF TRANSACTIONAL ITAND ANALYTICAL IT
As we just discussed, widespread implementation of ERP and e-commerce
systems offer the promise of homogeneous, transactional databases that will facilitate
managerial decision-making. In many companies, however, the scope and flexibility of
installed ERP systems has been less than desired or expected. New ERP systems that are
modular and Web-enabled are scarcely past the drawing board stage, but we can expect to
see significant improvements over the next three to five years.
In any event, competitive advantage cannot be gained simply through faster and
cheaper communication of data. As many managers have come to realize, ready access to
transactional data does not automatically lead to better decision making. In reality,
Enterprise Resource Planning is a misnomer because it fails to provide insights into
decisions affecting “resource planning.” The guiding principle for overcoming these
deficiencies is
To effectively apply IT in managing its supply chain, a company must distinguish
between the form and function of Transactional IT and Analytical IT
Transactional IT is concerned with acquiring, processing and communicating raw data
about the company’s supply chain, and with the compilation and dissemination of
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-12
reports summarizing these data. The data may originate from internal sources, such as a
general ledger system or a manufacturing process control system, or it may originate from
external sources, such as an order placed over the internet or trucking rates of a common
carrier accessed by EDI.
By contrast, Analytical IT evaluates supply chain planning problems using
descriptive and normative models. Descriptive models, such as demand forecasting or
managerial accounting models, describe how supply chain activities, costs, constraints and
requirements may vary in the future. Normative, or optimization, models, such as a linear
programming model for capacity planning, describe the space of supply chain options
over which the supply chain manager wishes to optimize his/her decisions. Normative
models are constructed from supply chain decision databases using descriptive models
and data aggregation methods. These decision databases are discussed in detail in
Chapter 6.
Analytical IT is not very dissimilar in meaning from the term decision support
system (DSS). We have avoided using this term because it has come to connote an
unsystematic application of ad hoc methods to the analysis of business decision
problems. The implication is that each new decision problem requires the design and
implementation of a new model and a new DSS. We take an opposite viewpoint by
arguing that optimization models provide a rigorous, rich and coherent discipline for
constructing and deploying general-purpose tools. These tools are the cornerstone of
Analytical IT for integrated supply chain management. Optimization models are
discussed in detail in Chapters 3, 4, and 5.
Differences between Transactional IT and Analytical IT can be contrasted across a
number of aspects. In the paragraphs that follow, we discuss six contrasting aspects.
Aspect: Time frame addressed
Transactional IT: Past and present
Analytical IT: Future
Transactional IT focuses on communicating, storing and reporting on data
describing the company’s current supply chain operations. These data are added to
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-13
historical databases. They may describe internal operations, such as the orders shipped
today from a distribution center, or the tons of product manufactured today by a machine
in a plant. They may also describe the environment in which the firm does business, such
as industry sales for the previous quarter. Analytical IT picks up where Transactional IT
leaves off by extrapolating data into the future, and analyzing one or more scenarios to
identify effective decisions.
Aspect: Purpose
Transactional IT: Communications
Analytical IT: Forecasting and decision making
As we just discussed, Transactional IT communicates data describing the
company’s current and past supply chain activities, while Analytical IT seeks to forecast
scenarios of the future and optimize decisions associated with these scenarios.
Uncertainties about the future depend on the length of the decision problem’s planning
horizon, and the nature of the industry in which the firm competes. The uncertainties
may be slight for short-term decisions such as the selection of routes for shipping
completed orders to customers over the next week. In such cases, operational plans may
safely be developed from a single scenario of the future. At the other extreme, strategic
plans stretching out five years, or more, may entail considerable uncertainty and require
evaluation of many scenarios using a model.
Aspect: Business Scope
Transactional IT: Myopic
Analytical IT: Hierarchical
By its nature, Transactional IT is myopically concerned with current transactions
and the compilation of histories based on them. Analytical IT addresses future decisions
through a hierarchy of decision problems at all levels of planning, operational, tactical and
strategic. Thus, in the short-term, it may address myopic operational decisions, while, in
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-14
the long-term, it may address global facility location and mission decisions that include
aggregate descriptions of these same operational decisions to be made at a much later time.
Aspect: Nature of databases
Transactional IT: Raw and lightly transformed objective data
Analytical IT: Raw, moderately and heavily transformed data that is both objective
and judgmental
The databases created by Transactional IT are derived from raw data that is stored
in formats that leave the data unchanged or “lightly transformed”. We use this expression
to define, admittedly in a vague way, the limits of Transactional IT. It refers to operations
on data that are easy to understand although, for large data sets, the resulting
transformations may require considerable processing time. An example of lightly
transformed data is a report based on aggregate product categories of the cost and volume
of SKU’s acquired last quarter by a retailing company. Another example is the
computation of average costs for shipping full truckloads last month from all company
sourcing points to all market zones. By contrast, optimization models employed by
Analytical IT require data inputs derived from raw data that may involve significant
transformations. For example, the mapping of general ledger costs at each production
plant into direct and indirect product and process costs, and indirect plant costs, all of
which may be fixed or variable, with variable costs that may be linear or nonlinear.
Databases created and used by Analytical IT tools will also contain judgmental
data about the company’s supply chain. An example might be constraints that mitigate
production risk by limiting any plant to making no more than 75% of next year’s
forecasted demand of certain products. Another example is a constraint limiting to 400
miles the distance from each market to the distribution center that serves it.
These differences between transactional and analytical databases reflect the
differences in business scope discussed above. From a bottom-up perspective, the
handoff from Transactional IT to Analytical IT occurs when the company seeks to
optimize operational plans for the short-term future. Data that is irrelevant to operational
decision making, such as addresses where invoices for customer orders should be mailed,
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-15
is stripped from transactional databases, while the remaining data are fed to appropriate
supply chain decision databases. Some modeling practitioners estimate that 80%, or more,
of a transactional database is irrelevant to decision making and model construction.
As the planning horizon of decision problems to be evaluated by Analytical IT
stretches further into the future, the link between Transactional IT and Analytical IT
becomes more complex. The broader scope of longer term decision problems requires
aggregations of transactional data to provide the model and the decision makers with a
better view of their “planning forest” rather than the “trees” representing details of the
company’s operations. The aggregations must be reversed when longer-term plans are
translated into operational plans.
Aspect: Response time for queries
Transactional IT: Real-time
Analytical IT: Real-time and batch processing
Computing speeds have reached a point where users expect instantaneous, or at
least very fast, responses to data queries. This is especially true for Transactional IT
responsible for retrieving raw data from corporate databases. Of course, some
applications involve databases that are so large that rapid response to queries requires a
network of dedicated computers. For example, to support system-wide queries about
inventory, large retailing companies stocking 50,000 SKU’s across 500 stores and 10
distribution centers employ a data decomposition scheme in which multiple computers,
each dedicated to a subset of the company’s product line, are accessed through a server.
For certain types of Analytical IT, such as a system producing forecasts of
demand for a single product by weeks over the coming year, the response may also be
nearly instantaneous. Although the underlying model may perform a non-trivial amount
of number-crunching in determining the forecast, computation is fast enough that the user
does not perceive an appreciable delay. For other types of applications, such as the
determination of a vehicle routing schedule for daily deliveries to 1000 customers, 15
minutes, or longer, may be required to generate and solve an optimization model on a
high-end PC. Considerable number-crunching in a batch mode is required for this
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-16
application because it entails comprehensive evaluation of a complex system of customers
and proposed routes, rather than myopic analysis of data pertaining to a single customer
or a single route.
The distinction between myopic analysis and comprehensive systems analysis of
supply chain decision problems and their requirements for response time, is not yet
widely understood or appreciated. Many problems, such as the vehicle routing problem
just cited, could be analyzed by myopic methods in real-time, or in no more than a few
seconds. For complex problems, however, the plans identified by myopic methods will be
markedly inferior to those determined by a global model and optimization method. If
he/she understood this difference, the distribution manager would certainly be willing to
wait several minutes to obtain a plan produced by a batch run of an optimization model
that delivers all orders using 10% fewer trucks. Moreover, considerable refinement by a
human analyst may be needed to make the solution produced by a myopic method
acceptable for implementation.
The importance of response time diminishes, but does not disappear, as the scope
of the supply chain problems to be analyzed moves from daily, operational concerns to
tactical and strategic ones. Analysis of tactical decisions affecting activities over the next
few days or months must still be made in a timely fashion. Strategic planning may require
evaluation of many scenarios to complete a study within a tight timeframe, thereby
constraining the response time available for extracting useful answers from an
optimization model describing a single scenario.
Aspect: Implications to business process redesign
Transactional IT: Substitute for or eliminate inefficient human effort
Analytical IT: Coordinate overlapping managerial decisions
The impact of IT on business process redesign is an immense subject that we
discuss briefly here. We will return to it in §12.6. At this juncture, we simply point out
that Transactional IT and Analytical IT have qualitatively different impacts on the
organization and management of a company’s supply chain. Transactional IT has allowed
communication of data describing operational business processes to be automated and
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-17
made more efficient. It has also provided managers with timely data to make better-
informed intuitive decisions about short-term operations, although some decisions could
be improved if they were based on model results.
Analytical IT allows supply chain decisions to be integrated across managerial
responsibilities, and across levels of planning, but to be fully exploited, it entails major
organizational change. Although such changes are underway in many companies, their
ultimate nature and extent depends on future IT and organizational developments about
which we speculate in Chapter 12. In summary, Analytical IT seeks to systematically
identify opportunities for improving the management of the company’s supply chain by
functional and inter-temporal integration of decisions, whereas Transactional IT addresses
myopic opportunities for such improvement.
2.4 HIERARCHY OF SUPPLY CHAIN SYSTEMS
In the previous section, we emphasized the importance of inter-temporal
integration of supply chain activities, as well as their functional and geographical
integration. Inter-temporal integration can be fully achieved only by the application of a
suite of modeling systems to the gamut of strategic, tactical and operational planning
decision problems faced by the company. These Analytical IT systems are linked to
overlapping supply chain decision databases created from data provided by Transactional
IT systems. Companies offering ERP software have realized this need. They are actively
expanding their offerings to include modeling systems for all levels of planning, either by
developing such systems themselves or by acquiring companies with modeling software.
Components of the Supply Chain System Hierarchy
In Figure 2.2, we display the Supply Chain System Hierarchy comprised
of six types of optimization modeling systems and four transactional systems responsible
for inter-temporal, functional, and geographical integration of supply chain activities in a
manufacturing and distribution company with multiple plants and distribution centers. As
shown in the Figure, the six types of optimization modeling systems are Analytical IT
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-18
Ent
erpr
ise
Res
ourc
e P
lann
ing
Sys
tem
Mat
eria
ls
Req
uire
men
tsP
lann
ing
Sys
tem
s
Dis
trib
utio
ns
Req
uire
men
tsP
lann
ing
Sys
tem
s
Sco
pe
Str
ateg
ic A
naly
sis
Lon
g-te
rmT
acti
cal A
naly
sis
Sho
rt-t
erm
Tac
tica
l Ana
lysi
s
Ope
rati
onal
Ana
lysi
s
Log
isti
cs
Opt
imiz
atio
nM
odel
ing
Sys
tem
Sup
ply
Cha
in S
yste
m H
iera
rchy
Fig
ure
2.2
Tac
tica
l O
ptim
izat
ion
Mod
elin
g S
yste
m
Str
ateg
ic
Opt
imiz
atio
nM
odel
ing
Sys
tem
Dem
and
For
ecas
ting
and
Ord
erM
anag
emen
t Sys
tem
Pro
duct
ion
Pla
nnin
g O
ptim
izat
ion
Mod
elin
g S
yste
ms
Pro
duct
ion
Sch
edul
ing
Opt
imiz
atio
n M
odel
ing
Sys
tem
s
Dis
trib
utio
n S
ched
ulin
g O
ptim
izat
ion
Mod
elin
g S
yste
ms
Ana
lyti
cal I
T
Tra
nsac
tion
al I
T
Ext
erna
l Dat
a M
anag
emen
t Sys
tem
s
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-19
and the four other systems are Transactional IT. Strictly speaking, the Demand
Forecasting and Order Management System is a hybrid with analytical capabilities for
forecasting demand and transactional capabilities for handling customer orders.
We have also shown an important linkage between external data management
systems maintained by the company’s customers and suppliers and the company’s
enterprise transactional data. Recent advances in e-commerce offer the promise to
streamline and enhance such communication. They also increase the need for modeling
systems, especially to support operational decision-making across multiple firms.
The Supply Chain System Hierarchy in Figure 2.2 is hypothetical. To the best of
our knowledge, no company has implemented and integrated all nine types of systems,
although many companies have implemented several of them. Moreover, the components
and structure of the Hierarchy may appear arbitrary and we would expect that they might
be modified for specific applications. Still, based on the author’s participation in scores of
projects in which supply chain modeling systems have been developed, the Hierarchy
represents the most likely configuration needed to analyze strategic, tactical and
operational supply chain planning problems in a firm that both manufactures and
distributes products.
The transactional and scheduling systems in the System Hierarchy represent the
bottom-up thrust in supply chain management. IT developments are the driving force for
innovations in this area, with business process re-design as a natural consequence. The
area is red hot with annual sales of software in the hundreds of millions of dollars and
growing rapidly.
Distinctions among the transactional and scheduling systems displayed in Figure
2.2 have become blurred. Software companies offering ERP systems have either acquired
or entered into alliances with companies offering operational modeling systems. Similarly,
some DRP systems include modules for vehicle scheduling and forecasting. For our
purposes here, we choose to maintain separation between the form and function of the
modeling and transactional systems.
In seeking to better manage their operations, a company must decide if it wishes
to acquire off-the-shelf systems, or to develop customized systems implemented by its
internal IT staff, by outside system developers, or some combination of the two.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-20
Customized systems, if they are implemented in an effective and timely manner, are more
likely to provide the company with competitive advantage than the acquisition of off-the-
shelf systems.
The number and diversity of off-the-shelf systems for operational supply chain
management is increasing. Most off-the-shelf scheduling systems to support MRP and
DRP decision-making use heuristics that are less formal, and less effective in identifying
demonstrably good operational plans, than the optimization models we envision for the
Production Scheduling and Logistics Optimization Modeling Systems. They rely heavily
on graphical user interaction in developing scheduling solutions, often placing too high a
burden on historical rules and the judgment of a human scheduler to extract a good
operational plan from a complex data set. We believe the unified optimization
methodology discussed and illustrated in Chapter 5 has the potential to provide modeling
systems with superior performance.
Strategic optimization modeling systems in the System Hierarchy reflect the top-
down thrust in supply chain management. The driving force is senior management's need
for strategic analysis in the face of globalization of the company's markets and supply
chains, and competition in cost and service. A typical strategic planning study is being
performed by consultants who employ an optimization modeling system. They exercise
the system to provide management with quantitative insights into the evolution and re-
design of their supply chains, and answers to "what if" questions about the long-term
future.
The study mode for applying these systems is often very useful, but short-
sighted. Since the strategic supply chain problems evaluated by the modeling systems
almost certainly will not disappear at the end of the study period, the company would
profit greatly by making modeling analysis a permanent part of its strategic review
processes. Nevertheless, senior management encounters several organizational barriers
when it tries to use a strategic modeling system on a regular basis.
First, the company must commit to the design and implementation of IT
procedures for collecting and updating the supply chain decision database. It must also
commit to the creation and training of analysts who will devote a significant portion of
their time in performing on-going strategic evaluations. Second, training entails a complete
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-21
transfer of the modeling system technology from the modeling system developers and, in
many cases, the external consultants who perform the study, to the company. Finally, the
company must implement processes whereby senior management works with analysts
and lower level managers in performing modeling studies, reviewing results, and
implementing plans that they suggest.
Long-term and short-term tactical supply chain planning have thus far been
mainly ignored by managers and consultants. They are the most difficult areas in which to
develop better planning methods, based in part on optimization modeling systems.
Despite the growing number of applications of strategic optimization modeling systems,
we have seen few initiatives to move down the Hierarchy to develop and use these
systems for related tactical planning problems. The lack of interest by companies that
have successfully used optimization modeling systems for strategic studies in extending
them to tactical modeling applications is frustrating for modeling practitioners. From a
technical perspective, such extensions are easy to accomplish because the model and the
supply chain decision database will be validated during the study.
Still, this reluctance is not surprising since repetitive use of a Tactical
Optimization Modeling System requires considerable business process re-design. Tactical
applications also require the development and upkeep of supply chain decision databases,
which, as we already observed, are not yet well understood. Despite the difficulties, we
are optimistic about the ultimate breakthrough of tactical modeling applications because
the potential rewards are so great. Limited applications have demonstrated that a
manufacturing or distribution company can expect to reduce its total supply chain costs
by 5%, or more, by implementing plans identified by a modeling system. The tool is also
valuable in helping management adjust to unexpected changes in its business environment,
such as a fire at a company plant or a strike at a key vendor.
Starting from the bottom-up, the following are synopses of the capabilities of each
system type.
Enterprise Resource Planning (ERP) System: The ERP System manages the
company’s transactional data on a continuous, real-time basis. This System standardizes
the company’s data and information systems for order entry, financial accounting,
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-22
purchasing, and many other functions, across multiple facilities and business units.
Despite the claim implied by the term ERP, effective “resource planning” across the
“enterprise” can be identified only by optimization models created using data from the
ERP System.13
Materials Requirement Planning (MRP) System: Analysis with the MRP System
begins with a master production schedule of finished products needed to meet demand
in each period of a given planning horizon. Using these data, along with a balance on
hand of inventories of raw materials, work-in-process and finished goods, a bill of
materials description of the company’s product structures, and machine production
data, the MRP System develops net requirements by period of raw materials and
intermediate products to be manufactured or ordered from vendors to meet demand for
finished products. Products at all stages of manufacturing are analyzed by the MRP
System at the SKU level.14
Distribution Requirements Planning (DRP) System: Analysis with a DRP System
begins with forecasts of finished products to be transported, a balance on hand of
inventories of these products at plants and distribution centers, and inventory
management data such as safety stock requirements, replenishment quantities, and
replenishment times. In conjunction with the Distribution Scheduling Optimization
Model Systems, the DRP System then schedules in-bound, inter-facility, and out-bound
shipments through the company’s logistics network, taking into account a wide range of
transportation factors such as vehicle loading and routing, consolidations, modal choice,
channel selection, and carrier selection. Products throughout the logistics network are
analyzed by the DRP System at the SKU level.15
Demand Forecasting and Order Management System: This System combines data
about current orders with historical data to produce requirements for finished products to
be met by the operational, tactical and strategic plans. For operational and short-term
tactical planning, an important challenge is to manage the transition from forecasts, which
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-23
have a significant degree of uncertainty, to orders, which have much less uncertainty.
Longer-term planning requires linkages to data on industry and economic factors that have
a high degree of uncertainty.16
Production Scheduling Optimization Modeling Systems: These are modeling
systems located at each plant in the company’s supply chain that address operational
decisions such as the sequencing of orders on a machine, the timing of major and minor
changeovers, or the management of work-in-process inventories. The models must fit the
environment, which may be discrete parts manufacturing, process manufacturing, job-
shop scheduling, or some hybrid. 17 A single facility may require different modeling
systems at different stages of manufacturing; for example, fine paper production at a mill
involves process manufacturing to produce mother rolls of paper followed by job-shop
scheduling to produce the final products.
Distribution Scheduling Optimization Modeling Systems: The manufacturing and
distribution company faces a variety of vehicle and other scheduling and operational
planning problems. In addition to local delivery of products to customers, some
companies must decide on a short-term basis which distribution center should serve each
market based on inventory availability. As with production scheduling, distribution
scheduling problems and models vary significantly across industries.18
Production Planning Optimization Modeling Systems: Each plant in the
company’s supply chain uses its version of this optimization modeling system to
determine a master production plan for the next quarter for each stage of manufacturing,
along with resource levels and resource allocations for each stage, that minimize
avoidable manufacturing costs. As part of the optimization, the model also determines
work-in-process inventories, major machine changeovers, and make-or-buy decisions.
The models used by this System will be multi-period as well as multi-stage. Therefore,
for reasons of computational necessity, products are aggregated into product families.
These aggregations are reversed when the System hands off the master schedule to the
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-24
plant’s Production Scheduling and MRP Systems. Although many papers have
appeared in the academic literature discussing production planning models with this
broad scope, few modeling systems based on them have yet been implemented.19
Logistics Optimization Modeling System: This System determines a logistics master
plan for the entire supply chain that analyzes how demand for all finished products in all
markets will be met over the next quarter. Specifically, it focuses on the assignment of
markets to distribution centers and other facilities responsible for sourcing them. Its goal
is to minimize avoidable transportation, handling, warehousing and inventory costs across
the entire logistics network of the company, while meeting customer service requirements.
Again, for reasons of computational necessity, finished products are aggregated into
product families and markets are aggregated into market zones. These aggregations are
reversed when the System hands off the master schedule to the plant’s Distribution
Scheduling and DRP Systems. This type of optimization modeling system has also not
yet been widely implemented.
Tactical Optimization Modeling System: This System determines an integrated
supply/manufacturing/distribution/inventory plan for the company’s entire supply chain
over the next 12 months. Its goal may to be minimize total supply chain cost of meeting
fixed demand, or to maximize net revenues if product mix is allowed to vary. Raw
materials, intermediate products and finished products are aggregated into product
families. Similarly, markets are aggregated into market zones. This is another type of
optimization modeling system that has not yet been widely implemented.20
Strategic Optimization Modeling System: This System is used to analyze resource
acquisition and other strategic decisions faced by the company such as the construction of
a new manufacturing facility, the break-even price for an acquisition, or the design of a
supply chain for a new product. Its goal may be to maximize net revenues or return on
investment. A number of off-the-shelf packages, with varying degrees of modeling
capabilities, are available for this type of application.21
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-25
FREQUENCY OF ANALYSIS, CYCLE TIMES AND RUN TIMES OFSUPPLY CHAIN SYSTEMS
In the following sub-sections, we discuss interactions among systems immediately
adjacent to one another in the Supply Chain System Hierarchy depicted in Figure 2.2.
Before delving into these details, we need to examine how and when these systems are
applied. To this end, Table 2.1 reviews several timing features of each system:
• frequency of analysis – the number of times each year, quarter or month that
managers and planners use the system
• planning time – how long it takes to complete analysis of the planning problems
with the system each time it is used
• run time – batch time required for each run of the system
The times shown in Table 2.1 are representative of the systems in the Hierarchy. They
may vary significantly from company to company. The frequency of analysis will be
much longer than once a week for the Tactical Supply Chain Modeling System, and much
shorter than once a quarter for the Production Scheduling Modeling System. The planning
horizons of the Modeling Systems, the MRP System, and the DRP System overlap. This
facilitates coordination and communication among them.
The column labeled Model Structure refers to the number of periods incorporated
in models generated by the modeling system. For example, a strategic optimization model
will typically be a one period, or snapshot model, where the period is one year. A
production planning model might be a six period model where the first four periods are
weeks and the final two periods are months.
As we descend in Table 2.1 from strategic to operational systems, the planning
horizon becomes shorter while the description of time in the model structures, as
measured by the number of periods in the models, becomes more detailed. In addition,
the objective function shifts from net revenue maximization to avoidable cost
minimization as we move from strategic to operational planning. Although, net revenue
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-26
Feat
ures
of
Ana
lytic
al a
nd T
rans
actio
nal S
yste
ms
Tab
le 2
.1
Plan
ning
Hor
izon
Mod
elSt
ruct
ure
Obj
ectiv
e Fu
nctio
nR
unT
ime
Stra
tegi
cO
ptim
izat
ion
Mod
elin
gSy
stem
1 -
5ye
ars
Yea
rly
snap
shot
s
max
imiz
e ne
t re
venu
esor
ret
urns
on
asse
ts
1 0
- 60
min
Tac
tical
Opt
imiz
atio
nM
odel
ing
Syst
em
12 m
onth
s3
mon
ths,
3 qu
arte
rs
min
imiz
e to
tal c
ost o
f m
eetin
gfo
rcas
ted
dem
and
orm
axim
ize
net r
even
ue b
y va
ryin
gpr
oduc
t mix
60 -
120
min
Prod
uctio
nPl
anni
ng
Opt
imiz
atio
n M
odel
ing
Syst
em
13 w
eeks
4 w
eeks
,
2 m
onth
s
min
imiz
e av
oida
ble
prod
uctio
n an
d in
vent
ory
cost
s10
- 3
0m
in
Log
istic
s O
ptim
izat
ion
Mod
elin
g Sy
stem
13 w
eeks
4 w
eeks
,
2 m
onth
s
min
imiz
e av
oida
ble
logi
stic
s co
sts
10 -
30
min
Fore
cast
ing
and
Ord
erM
anag
emen
tSy
stem
1 w
eek
-
5 ye
ars
Var
ied
not a
pplic
able
10 s
ec -10
min
Ent
erpr
ise
Res
ourc
ePl
ainn
g Sy
stem
not a
pplic
able
Rea
l-tim
e or
Con
tinuo
usno
tap
plic
able
DR
PSy
stem
7 da
ysto
28 d
ays
not a
pplic
able
60 m
in7
days
to28
day
s
MR
PSy
stem
sno
t app
licab
le60
min
7 da
ysto
28 d
ays
7 da
ysto
28 d
ays
Dis
trib
utio
n Sc
hedu
ling
Opt
imiz
atio
n M
odel
ing
Syst
ems
10 m
in7
days
to28
day
s
7 da
ysto
28 d
ays
Prod
uctio
nSc
hedu
ling
Opt
imiz
atio
n M
odel
ing
Syst
ems
min
imiz
e m
yopi
c pr
oduc
tion
cost
s
Freq
uenc
y of
Ana
lysi
s
Onc
ea
year
Onc
ea
mon
th
Onc
ea
wee
k
Onc
ea
wee
k
Var
ied
Onc
ea
wee
k
Onc
ea
wee
k
Onc
ea day
Onc
ea day
Plan
ning
Tim
e
1 -
2m
onth
s
1 w
eek
1 da
y
1 da
y
Var
ied
1 -
3 h
ours
1 -
3 h
ours
30 m
in
30 m
in10
min
7 da
ysto
28 d
ays
7 da
ysto
28 d
ays
min
imiz
e m
yopi
c di
stri
butio
n co
sts
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-27
maximization should be sought at all planning levels, the company may have few options
to affect revenue at the operational level.
One can expect or hope that, in the coming years, net revenue maximization will
work its way down the Hierarchy as the company’s management improves its abilities to
integrate supply chain and demand management decisions. For example, the Production
Scheduling Modeling System could be employed to maximize short-term net revenues by
identifying which customized orders to accept or reject, or to determine prices for such
orders so as to guarantee healthy margins. Such a change in using this System would
require changes in business processes to support both the requisite analysis and
negotiations with customers.
COMMUNICATION AMONG SUPPLY CHAIN SYSTEMS OF DATA AND
DECISIONS
In the paragraphs that follow, we discuss interactions among the systems in the
Supply Chain System Hierarchy. In effecting these interactions, decisions determined by
the Modeling Systems become input data to other systems with which they
communicate.
ERP, MRP, DRP and Forecasting and Order Management Systems
Figure 2.3 depicts interactions among the ERP, MRP, DRP and Forecasting and
Order Management Systems. Although we have shown them as separate systems, they
could be viewed as a single ERP entity dedicated to acquiring, communicating and managing
transactional data requirements across the company. The MRP and DRP Systems that are
one level up from the ERP System develop and disseminate detailed production and
distribution schedules. A separate MRP System is employed in each plant, whereas the
DRP System addresses distribution operations across the entire company. These Systems
are mainly transactional programs that translate master production and distribution
schedules into detailed schedules. The Systems also keep track of actual production and
distribution data. The typical planning horizon for these schedules is 7 to 28 days.
The ERP system provides the MRP and DRP systems with detailed data about
costs, capacities and equipment. It also passes data about orders to the Forecasting and
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-28
Operational Supply Chain Systems
Figure 2.3
distributionmanagers
salesmanagers
Forecasting and Order
ManagementSystem
MaterialsRequirements
PlanningSystem
DistributionRequirements
PlanningSystem
EnterpriseResource Planning System
productionmanagers
CostsCapacitiesMachine
performancedata
Detailedproductionschedules
CostCapacitiesVehicle
performancedata
Detailedtransportation
schedules
Detailedfinished goods inventories at distribution
centers
Orders
Detailedwip and
finished goods inventories at
plants
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-29
Order Management System, which in turn passes orders and forecasts to the MRP and
DRP Systems. The company’s production and distribution managers use detailed
schedules developed by these Systems to execute the company’s operational plans.
These data, along with data about inventories are also passed to the ERP System for
tracking, accounting and control purposes.
MRP and Production Scheduling Modeling Systems
Without the Production Modeling System, users of the MRP System must
determine master schedules and available capacities in an ad hoc way based on historical
rules-of-thumb. Although the typical MRP System has rudimentary tools intended to
assist company planners in determining schedules, they leave much to be desired. For
example, it might compute capacity loadings implied by the master schedule, but cannot
adjust the master schedule if the loadings exceed available capacity.
In short, the MRP System cannot identify a short-term schedule, required resource
levels, and their allocations, which minimize total operational costs over the short-term
planning horizon. Moreover, it cannot assist schedulers in determining a feasible schedule,
or which orders to delay, when manufacturing capacity is tight. As a result, in using the
MRP System without the Production Scheduling Modeling System, production managers
can only muddle through the scheduling process by using trial-and-error methods. For this
reason, the company needs a Modeling System that employs optimization models and
methods to determine an effective production schedule over a 13 week planning horizon,
with particular attention paid to the next 4 weeks, which span the 28 day horizon of the
MRP system. The typical model generated by this System looks out 13 weeks to ensure
stability to the detailed plan for the next 28 days that ultimately will be executed according
to the MRP System.
As shown in Figure 2.4, the optimization model determines production set-ups,
production runs, discretionary resource levels, work-in-process and finished goods
inventories so as to minimize avoidable costs associated with attempting to meet
customer orders. We say “attempting” to meet customer orders because the company
may encounter order schedules that cannot be met. In such an event, based on implicit and
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-30
Linkages between Short-term Tacticaland Operation Production Planning Systems
in a Plant
Figure 2.4
By production family: Master production schedule Wip inventories Finished goods inventories at plantBy work centers: Resource levels Resource allocationsOrders to be backlogged
ProductionSchedulingModelingSystem
Aggregation Disaggregation
MaterialsRequirements
PlanningSystem
Detailed production, cost and inventory dataDetailed order and demand dataSemi-permanent data about machine performance and labor requirements
productionplanner
productionmanagers
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-31
explicit penalties associated with late deliveries, the production scheduling model assists
production managers in determining which orders will be completed and shipped late.
The links between the Production Scheduling Modeling System and the MRP
System involve aggregation when data are fed upward from the MRP System to the
Modeling System, and disaggregation when data are fed downward from the Modeling
System to the MRP System.22 Upward aggregation entails aggregation of products into
products families and detailed time dependent data, such as scheduled maintenance or
machine changeovers, into aggregate time dependent data, such as the week in which these
events will take place. Downward disaggregation entails translation of production
schedules and inventories of product families into details regarding individual products. It
also entails translation of the aggregate timing of time-sensitive decisions into more
detailed timing. The disaggregation transformation is essentially an inversion of the
aggregation transformation with rules-of-thumb applied to ensure that the resulting details
are efficient and best satisfy downstream production and customer requirements. Thus, in
designing the upward aggregation, care must be taken to ensure that the corresponding
downward disaggregation can be easily and accurately carried out.
DRP and Logistics Modeling Systems
Figure 2.5 shows the relationship between the DRP System and the Logistics
Modeling System. Unlike the production systems just discussed that separately analyze
each plant, these Systems analyze decisions across the company’s entire logistics network,
which might include several plants and distribution centers, and several hundred markets.
They also coordinate company transportation activities with those of the vendors.
Otherwise, the company’s motivation for implementing and deploying the Logistics
Modeling System is the same as that for the Production Scheduling Modeling System.
Without such a System, distribution managers using the DRP System must muddle through
the short-term scheduling of transportation movements and the operations of distribution
centers to support them. For example, the DRP system has the capability to heuristically
optimize daily vehicle loading and routing decisions, but cannot determine which
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-32
Linkages between Short-term Tacticaland Operation Distribution Planning Systems
Figure 2.5
By production family: Master transportation schedule Finished goods inventories at DC'sFor each DC: Operating schedule Resource levels Resource allocations
LogisticsModelingSystem
Aggregation Disaggregation
DistributionRequirements
PlanningSystem
Detailed transportation capacity and cost dataDetailed inventory dataDetailed costs and capacity data for DC's
logisticsmanagers
logisticsmanagers
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-33
distribution centers should serve each market, and how operations at the distribution
centers should be scheduled, so as to minimize short-term costs.
The Logistics Modeling System determines a master transportation schedule that
includes in-bound shipments of raw materials and parts to the plants, inter-plant shipments
of intermediate and finished products, shipments of finished products to distribution
centers, and out-bound shipments to the markets of finished products. Decisions about the
latter shipments fine-tune the longer-term assignment of markets to distribution centers
determined by the Tactical Modeling System. In addition, the Logistics Modeling System
makes modal choices for large shipments based on timing considerations; for example, a
choice between a single large rail shipment from a plant to a distribution center, or many
truck movements spread out over a month’s time.
Production Scheduling Modeling Systems, Logistics Modeling System, TacticalModeling System, Demand Forecasting and Order Management System
The Tactical Modeling System is the lowest level system in the Hierarchy that
analyzes decisions across the company’s entire supply chain. As shown in Figure 2.6, it
passes aggregate details about the optimal supply chain plan for each of the three months
of the immediate quarter to the Production Scheduling Modeling Systems, one in each
plant, and to the Logistics Modeling System. The details of this plan are disaggregated to
provide guidelines for the Production Scheduling Modeling System and the Logistics
Modeling System. Disaggregation may entail refinement of product families and the
timing of resource planning decisions. Schedules developed by the lower level systems are
fed back to the Tactical Supply Chain Modeling System by reversing these
disaggregations. These schedules reflect short-term commitments that the higher level
system treats as fixed and given.
Unlike the interactions discussed previously, we have shown linkages, which are
two directional, between these Modeling Systems and the Demand Forecasting and Order
Management System. In particular, the Modeling Systems receive order and forecasting
information from this System, while the Tactical Supply Chain Modeling System sends
suggested product mix strategies to this System. Marketing and sales personnel can use
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-34
Lin
kage
s A
mon
g T
actic
al M
odel
ing
Syst
ems,
Prod
uctio
n Sc
hedu
ling
Mod
elin
g Sy
stem
s, L
ogis
tics
Mod
elin
g Sy
stem
san
d D
eman
d Fo
reca
stin
g an
d O
rder
Man
agem
ent S
yste
ms
Figu
re 2
.6
Dem
and
Fore
cast
ing
and
Ord
erM
anag
emen
t Sys
tem
Ord
ers
and
Fore
cast
s
Prod
uct M
ixSt
rate
gies
Tac
tical
Mod
elin
g Sy
stem
Log
istic
sM
odel
ing
Syst
em
Prod
uctio
n Sc
hedu
ling
Mod
elin
g Sy
stem
s
Plan
t N
Prod
uctio
n Sc
hedu
ling
Mod
elin
g Sy
stem
s
Plan
t 1
Ass
ign
fore
cast
ed
dem
and
to p
lant
sR
esou
rce
Plan
ning
Inte
r-fa
cilit
y sh
ipm
ents
Ass
ignm
ent o
f m
arke
ts
to D
C's
and
pla
nts
Ass
ignm
ent o
f ve
ndor
s to
pla
nts
Tar
get i
nven
tori
es b
y fi
nish
ed p
rodu
ct f
or
each
per
iod
Det
ails
abo
utsh
ort-
term
sche
dule
sD
etai
ls a
bout
shor
t-te
rmsc
hedu
les
disa
ggre
gate
aggr
egat
e
disa
ggre
gate
aggr
egat
e
disa
ggre
gate
aggr
egat
e
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-35
these strategies in revising their plans so as to enhance the company’s projected net
revenues over the coming year.
Strategic Modeling System, Tactical Supply Chain Modeling System, DemandForecasting and Order Management System
The Strategic Modeling System assists senior management in determining the most
effective long-term configuration of the company’s supply chain network. Its models
analyze decisions about major resource acquisitions and divestments, and the manufacture
and distribution of new and existing products over the coming years. The implications of
these decisions to next year’s tactical plans are passed to the Tactical Modeling System
as depicted in Figure 2.7. Such data might include new facilities that will be available or
products to be manufactured, distributed and sold during that time frame. The Tactical
Modeling System provides detailed feedback to the Strategic System about how these
facilities will be used and how market demand will be met over the first year of a strategic
planning horizon.
The Demand Forecasting and Order Management System provides medium and
long term demand forecasts to the Tactical and Strategic Supply Chain Modeling
Systems. Conversely, the Strategic Supply Chain Modeling System provides the Demand
Forecasting System with feedback about the profitability of existing and new product
lines. This information can be used to develop marketing strategies for increasing sales of
profitable products. In fact, the Demand Forecasting System might well be extended to
include marketing models for achieving this end.23
Balancing Centralized and De-Centralized Decision-Making
An important underlying purpose of the System Hierarchy is to resolve
management's conundrum of wishing to make supply chain decisions in both a centralized
and a de-centralized manner24 Centralized decision-making is needed to realize efficiencies
stemming from integration. De-centralized decision-making is needed for rapid, detailed
execution of operations. As we discussed extensively above, the conflict can be resolved
by passing guidelines based on centralized planning using a Modeling System to a lower
level Modeling System.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-36
Demand Forecastingand Order
ManagementSystem
StrategicSupply Chain
Modeling System
TacticalSupply Chain
Modeling System
Aggregation Disaggregation
Details ofnext year's
strategy
Forecasts
Productmix
strategies
Supply chain network configuration
Major resources
New product strategies
Forecasts
Productmix
strategies
Linkages Among Strategic Supply Chain Modeling System,Tactical Chain Modeling System and Demand Forecasting and
Order Management Systems
Figure 2.7
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-37
For example, the supply chain manager uses the Tactical Optimization Modeling
System to determine short-term production targets for each plant. These targets are
passed as inputs to the Production Planning Optimization Modeling Systems, one for
each plant, which product managers use to determine more detailed plans, including a
master schedule and optimal capacity levels, for the plant to follow over the next quarter.
These plans in turn are passed to the Production Scheduling and MRP Systems, which
lower level managers use to determine a detailed implementation plan for the next month.
In addition, the lower level systems provide feedback to the high level systems about
necessary adjustments to the centralized plans made necessary by the realities of more
detailed operations.
2.5 LEGACY SYSTEMS AND LEGACY THINKING
Legacy planning systems are outdated computer systems passed on to IT
personnel and managers who employ artistry in trying to apply them to planning
problems that have changed, sometimes using awkward data linkages to new systems.
ERP systems allow a company to replace inefficient legacy systems as well as to
homogenize and integrate disparate corporate databases. Our interest here is to discuss
issues connected with replacing legacy modeling systems and improving legacy thinking
about how supply chain decisions should be made. While the effort required to replace a
legacy modeling system may be significant, overcoming the barriers due to legacy thinking
about supply chain decision-making is often a more difficult task.
In some instances, the legacy modeling system to be updated resides on a
mainframe computer. The company wishes to replace it with a modeling system residing
on a PC, which requires implementing programs that download data and upload plans
identified by the system. Such was the case for a company that manufacturers and
distributes food products that sought recently to replace a legacy system constructed in
the 1970’s. The system used mainframe modeling generation and optimization software
that had been successfully applied for over twenty years. Nevertheless, the company
wished to replace it because model generation time on the mainframe, due to inefficient
data acquisition programs and the need to share CPU time with other users was excessive.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-38
Tests with several PC-based, off-the-shelf modeling systems indicated that the
legacy system could easily be replaced by a better performing system. However, the
company showed no interest in expanding the scope of analysis provided by the model,
which, by 1970’s standards was sophisticated, but by those of the 1990’s, was
simplistic. Specifically, the legacy model sought to minimize the total cost of meeting
demand by assigning production to plants, making inter-plant shipments as needed, and
by sole sourcing markets with shipments from a unique plant. Descriptions of production
costs, capacities and transformation activities in the legacy model were simplistically
described by unit costs for each product at each plant and overall plant capacity. The
legacy model did not address fixed costs, capacity planning and economies of scale
associated with each of several important stages of manufacturing at each plant.
Moreover, it did not address decisions regarding the installation of additional capacity, or
the retirement of excess capacity, at new or existing plants.
In short, over the course of 20 years, use of the legacy system induced legacy
thinking in the company about integrated planning of its supply chain. For the reasons
just indicated, the legacy model produced plans that were probably seriously sub-
optimal. No one in the company was motivated to question current processes or to spend
time collecting data and making model runs to evaluate potentially better ways to manage
its supply chain. It is telling that the legacy system had been designed and implemented
by a corporate operations research group that gradually disappeared from the company,
leaving operations personnel without internal resources for evaluating new decision
processes and models. The operations managers could have sought outside help, but it
was unclear which software vendors or consultants to trust. Moreover, an exercise to
evaluate better models and modeling systems appeared expensive, although an improved
model and modeling system would have paid for itself several times over in the first year
of use.
As another example, a pharmaceutical company contacted a modeling practitioner
because it wished to model a critical production planning step in the manufacture of a
very successful product, with sales in the hundreds of millions of dollars, made from a
natural ingredient of varying quality. The step involving blending the natural ingredient to
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-39
produce the basic component used in subsequent manufacturing steps. Approximately 40
blends per year are made.
The company sought a modeling system that would automate and optimize a
planning process that was previously performed manually by a key employee. According
to legacy thinking about how to plan the blends, which we will describe very shortly, this
employee did an excellent job. Nevertheless, the company wished to automate the
process so that others could perform it; they also believed that it could be improved by
computer-based optimization.
Because yearly planning of the blends of natural ingredient had been done
manually, the goal in selecting them was to take the available pool of the natural ingredient
and try to find the set of blends that yielded the most uniform product. This problem
resembles blending of petroleum products, with the added qualitative concern that
product quality would be closely scrutinized by the Federal Drug Administration. In
addition, the blending constraints were defined relative to the composition of the pool,
rather than absolute constraints, which are imposed on petroleum products.
Since the natural ingredient is expensive to acquire, the practitioners suggested that
the company should expand the analysis to more carefully decide upon the amount of its
yearly purchases and more carefully control its inventories. In addition, rather than select
all 40 blends for the product at the beginning of each year, it was suggested that
production plans for each month could be modified according to marketing requirements.
Moreover, within limits, the optimization could select the pool of the natural ingredient
to be used for a given year as well as to optimize the blends made from it. These
suggestions fell on deaf ears. As a result, a model was designed and implemented that did
no more than optimize the manual process. The company was very pleased with the
resulting system and the blends it produced, which were superior to those produced by
hand.
The examples just cited of a company’s reluctance to go beyond legacy thinking in
managing its supply chain, are not exceptional. Although we could speculate further about
the reasons for this reluctance, we must recognize that it is still early days for the
development and use of modeling systems. Many companies are still grappling with ERP
system developments that they feel must proceed the development of analytical tools.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-40
Moreover, we must continue to educate students and managers about the value of rational
decision-making and the form, function, and benefits of models and modeling systems.
2.6 FINAL THOUGHTS
The growth in IT investment depicted in Figure 2.1 suggests that the information
revolution accelerated significantly in the 1990’s. Looking at the figure, one is also
tempted to conclude that the revolution hardly began before 1980. Unfulfilled
expectations about ERP systems and e-commerce indicate that developers and
consultants are still struggling with advances in software and business process redesign
needed to foster efficient and flexible systems for the purposes of Transactional IT. At
the same time, because managers have begun to recognize the need for Analytical IT, ERP
system companies are actively seeking to add optimization modeling systems to their
suite of offerings. Our discussion in this chapter is intended to set the stage for an in-
depth examination of optimization models and modeling systems that serve as the
“brains” of Analytical IT systems for supply chain management.
Furthermore, we recommend that Transactional IT developers pay more attention
to data requirements induced by modeling systems. New software is needed for creating
and applying supply chain decision databases that sit between corporate databases and
modeling systems. Transactional data that is irrelevant to decision-making should be
separated from transactional data that is relevant. Relevant data may require descriptive
analysis transformation before it can be used directly in an optimization model.
Despite 50 years of study in operations research models and methods, plus
numerous examples of successful modeling system implementations, we are still in the
early days of applying them in a pervasive and enduring manner to a range of supply
chain applications. As we shall attempt to demonstrate, operations research academics
and practitioners have filled their intellectual warehouse with models and methods that
offer great promise. The opportunity to pick good ideas from this warehouse and apply
them is exciting. At the same time, given the red hot pace of IT developments, modeling
practitioners must strive to fend off and displace mediocre solutions that, in the confusion
of the IT revolution, are being oversold to managers.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-41
2.7 EXERCISES
In addition to the exercises given below, modeling exercises involving data files and
discussion exercises involving white papers may be found on the book’s web site.
1. Inventory management is inherently an operational planning problem involving
decisions about when to order replacement stock and how much to order when such
decisions are made. Discuss reasons and situations in which inventory management is
also a tactical or strategic planning problem. In addition, discuss ways that inventory
decisions at the three levels of planning, operational, tactical, and strategic, are linked.
2. For a firm that manufactures industrial products, such as industrial chemicals or
printed circuit boards, describe conditions when it is appropriate to pursue
operational plans that maximize net revenue. Your discussion should include reference
to the operational time frame for such decision making, and to processes for
implementing plans that seek to maximize net revenues.
3. In their book, Reengineering the Corporation, Hammer and Champy state the
following25
“To recognize the power inherent in modern information technology and to
visualize its application requires that companies use a form of thinking that
businesspeople usually don’t learn and with which they may feel uncomfortable.
Most executives and managers know how to think deductively. That is, they are
good at defining a problem or problems, then seeking and evaluating different
solutions to it. But applying information technology to business reengineering
demands inductive thinking – the ability to first recognize a powerful solution
and then seek the problems it might solve, problems the company probably
doesn’t even know it has.”
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-42
a) According to the dictionary, deduction is the process of reasoning in which a
conclusion follows necessarily from the premises; reasoning from the general to
the specific.26 Moreover, induction is the process of deriving general principles
from particular facts or instances.27 Would you say that these terms were used
correctly in the above statement? Support your answer.
b) Provide arguments citing examples that support the intention of the statement
that management must explore opportunities for reengineering the corporation
to fully exploit new information technology.
c) Provide arguments citing examples where (so-called) inductive approaches have
proven counterproductive.
d) Provide a summary describing the extent to which you agree or disagree with the
statement.
4. In his book, A Primer on Decision Making, March describes rational decision
making as based on answers to four questions28
i) The question of alternatives: What actions are possible?ii) The question of expectations. What future consequences might follow
from each alternative? How lilkely is each possible consequence,assuming that alternative is chosen?
iii) The question of preferences: How valuable (to the decision maker) arethe consequences associated with each of the alternatives?
iv) The question of the decision rule: How is a choice to be made fromamong the alternatives in terms of the values of their consequences?
Proponents of rational decision making as a guiding force in managing companies
and organizations have modified their original thinking to one of bounded
rationality. They still believe that decision makers intend to be rational, but
most decision makers are limited by their mental capacities and the accuracy and
completeness of the information they have gathered. Specifically, they face
serious limitations in attention, memory, comprehension, and communication.
a) To what extent do models actualize and mechanize the theory of rational
decision making?
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-43
b) In your opinion, has the information revolution, including the advent of ERP
systems and e-commerce, relieved or exacerbated the limitations on making
rational decisions?
c) How can descriptive and normative models be used to overcome human
limitations of attention, memory, comprehension, and communication?
5. In section 1.3, we discussed Porter’s value chain and remarked that the
intersection of primary and support activities displayed in Figure 1.3 suggest
the need for a new type of matrix organization based on data and models.
Elaborate on this observation with particular reference to the hierarchy of
supply chain systems and the managers who use them discussed in section 2.3.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-44
2.7 NOTES
1 Lohr [1997].
2 This figure is from Lohr [1999] who examines Internet developments through the endof 1999. He suggests that “..it is probably too early to judge whether an Internetrevolution is truly under way. Historians say the Internet should be viewed mainly asthe latest advance in communications, a successor to the telegraph and the telephone,more a technological step than a leap forward.”
3 Vossen [1992] provides a tutorial of database management principles and a historicalperspective on developments up to the early 1990’s.
4 Robinson and Dilts [1999] give an overview of developments in ERP systems with anemphasis on the role that operations research models can play in extendingthesesystems to analyze supply chain decisions.
5 These figures are quoted in Deutsch [1998] who describes the pain that manycompanies have felt in trying to implement ERP systems.
6 Limitations as well as benefits of ERP systems are discussed in more detail byRobinson and Dilts [1998].
7 Latamore [1999] reports on developments in ERP systems, including extensions toWeb-enabled versions.
8 Canedy [1998] reviews business-to-consumer developments including projections oftotal Internet business in several markets by the year 2002, which are generally quitesmall relative to store sales. Of course, Internet developments far beyond 2002 arestill shrouded in considerable mystery.
9 See Tedeschi [1999].
10 Foster [1999] quotes this figure, attributing it to Forrester Research in Cambridge,Massachusetts.
11 See Foster [1999, 20].
12 Economist [2000] contains a discussion of the pros and cons of achieving perfectlyefficient markets over the Internet.
13 Starting in 2000, SAP, the leading ERP system company, will offer a suite ofmodeling modules to complement its ERP modules. Two other of the top five ERP
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-45
companies, J. D. Edwards and Baan, acquired smaller companies with supply chainmodeling systems.
14 For more information about MRP Systems, see Baker [1993] and Sipper and Bulfin[1997; 337-363].
15 For more information about DRP Systems, see Stegner [1994].
16 Demand forecasting is discussed in detail in § 6.7.
17 Shapiro [1993] reviews a variety of optimization models for production scheduling. Aproduction scheduling model and solution methodology is proposed in detail in § 5.3and § 5.5.
18 See Golden and Assad [1988] or Crainic and Laporte [1998] for a broad treatment ofvehicle routing algorithms and applications, and Hall and Partyka [1997] for a surveyof off-the-shelf packages for vehicle routing. A vehicle routing model and solutionmethodology is proposed in detail in § 5.2 and § 5.4.
19 Thomas and McClain [1993] provide a comprehensive literature survey of productionplanning models through the early 1990’s. Two examples of implemented productionplanning optimization modeling systems are those developed at Harris Corporation(Leachman et al [1996]) and Sadia (Taube-Netto [1996]). The applications at HarrisCorporation are discussed in detail in §10.5.
20 Tactical supply chain models and modeling systems are discussed in Chapters 7 and8.
21 Strategic supply chain models and modeling systems are discussed in Chapters 7 and8.
22 Graves [1982] gives an example of an optimization model for which aggregation anddisaggregation between short-term tactical and operational production scheduling arerigorously defined. For most applications, such rigor might be difficult to achieve.Nevertheless, practitioners would do well to push formal models and methods beforeresorting to ad hoc methods.
23 The integration of marketing and supply chain models is discussed in Chapter 8.
24 The notion that IT promotes schemes for simultaneous centralized and de-centralizedplanning in the firm was suggested by Hammer and Champy [1993; 93], althoughthey are vague about how it could actually be accomplished.
25 See Hammer and Champy [1993, 84].
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-46
26 See The American Heritage College Dictionary [1993, 362].
27 See The American Heritage College Dictionary [1993, 693].
28 See March [1994, 2-3].
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-47
2.8 REFERENCES
Baker, K. R. [1993], “Requirements Planning,” Chapter 11 in Handbooks in OperationsResearch and Management Science: Logistics of Production and Inventory, edited by S. C.Graves, A. H. G. Rinnoy Kan, P. H. Zipkin, North-Holland.
Canedy D. [1998], “Need Aspargus? Just Click It,” New York Times, C1, September 10,1998.
Crainic, T. G. and G. Laporte, editors [1998], Fleet Management and Logistics, KluwerAcademic.
Deutsch, C. H. [1998], “Software That Can Make a Grown Company Cry,” New YorkTimes, November 18.
Economist [2000], “How to be Perfect,” 82, February 12.
Foster, T. A. [1999], “Global eProcurement Solutions,” Supply Chain ManagementReview Global Supplement, 19-22, Spring.
Golden, B. L. and A. A. Assad, editors, Vehicle Routing: Methods and Studies, North-Holland, 1988
Graves, S. C. [1982], ‘Using Lagrangean Techniques to Solve Hierarchical ProductionPlanning Problems,” Management Science, 28, 260-275.
Hall, R. W. and J. G. Partyka, "On the Road to Efficiency," OR/MS Today, 24 (1997), 3,38-47.
Hammer, M. and J. Champy [1993], Reengineering the Corporation, HarperBusiness.
Latamore, G. B. [1999], “ERP in the New Millennium,” APICS, 9, No. 6, 28-32.
Leachman, R. C., R. F. Benson, C. Liu and D. J. Raar, "IMPReSS: An AutomatedProduction-Planning and Delivery-Quotation System at Harris Corporation-Semiconductor Sector," Interfaces, 26 (1996), 1, 6-37.
Lohr, S. [1997], “Information Technology Field is Rated Largest U.S. Industry,” NewYork Times, November 18.
Lohr, S. [1999], “The Economy Transformed, Bit by Bit,” New York Times, December20.
Robinson, A. G. and D. M. Dilts [1999], “OR & ERP,” ORMS Today, 26, No. 3, 30-35.
© Copyright 2000, Jeremy F. Shapiro. All rights reserved. 2-48
Shapiro, J. F. [1993], "Mathematical programming models and methods for productionplanning and scheduling," Chapter 8 in Graves, S. C., A. H. G. Rinooy Kan and P. H.Zipkin, North-Holland.
Sipper, D. and R. L. Bulfin, Jr. [1997], Production: Planning, Control, Integration,McGraw-Hill.
Stenger, A. J. [1994], “Distribution Resource Planning,” Chapter 17 in The LogisticsHandbook, edited by J. F. Robeson and W. C. Copacino, The Free Press.
Taube-Netto, M., "Integrated Planning for Poultry Production at Sadia," Interfaces, 26(1996), 38-53.
Tedeschi, R. [1999], “E-Commerce Report,” New York Times, C4, September 27, 1999.
Thomas, L. J. and J. O. McClain [1993], "An Overview of Production Planning," Chapter7 in Handbooks in Operations Research and Management Science: Logistics of Productionand Inventory, edited by S. C. Graves, A. H. G. Rinnoy Kan, P. H. Zipkin, North-Holland.
Vossen, G. [1992], “Databases and Database Management,” Chapter 4 in Handbooks inOperations Research and Management Science – Volume 3: Computing, edited by E. G.Coffman, Jr., J. K. Lenstra, A. H. G. Rinooy Kan.